Bengaluru-based artificial intelligence startup Sarvam AI has raised 400 million dollars in a landmark Series B funding round, making it the largest single AI investment in Indian startup history. The round was led by Lightspeed Venture Partners and Peak XV Partners, with participation from Khosla Ventures, Qualcomm Ventures, and the National Investment and Infrastructure Fund. The funding values Sarvam AI at 2.1 billion dollars, making it India's newest AI unicorn.
The Mission: AI for Every Indian
Sarvam AI was founded in 2023 by Dr. Vivek Raghavan and Dr. Pratyush Kumar, both former researchers at AI4Bharat — the IIT Madras initiative that has been building open-source AI tools for Indian languages since 2020. Their vision is audacious: build a large language model that is as capable as GPT-4 but optimized for India's linguistic and cultural context, supporting all 22 scheduled languages of the Indian Constitution.
The problem they are solving is fundamental. India has 1.4 billion people, but only approximately 10% speak English fluently. The remaining 90% — over 1.2 billion people — are largely excluded from the benefits of current AI systems, which are predominantly English-centric. A Hindi speaker in rural Bihar, a Tamil speaker in Chennai, or a Bengali speaker in Kolkata cannot effectively use ChatGPT, Gemini, or Claude for complex tasks in their native language. Sarvam AI aims to change this entirely.
Technical Approach
Sarvam's model architecture is based on a modified transformer design that incorporates linguistic features specific to Indian languages, including complex morphology, script diversity, and code-switching patterns common in Indian speech. The model is trained on a dataset of over 10 trillion tokens of Indian language content, including digitized books, news archives, government documents, social media, and specially commissioned content created by native speakers.
The company has developed proprietary techniques for handling the unique challenges of Indian languages, including the lack of standardized spelling in many languages, the prevalence of transliteration (writing Indian languages in Roman script), and the complex honorific systems that require understanding social context to generate appropriate responses. These innovations have resulted in benchmark performance that significantly exceeds existing multilingual models on Indian language tasks.
Products and Revenue
Sarvam AI has already launched several commercial products built on its foundation model. Sarvam-1, its flagship API, is being used by over 500 companies including HDFC Bank, Reliance Jio, and several state governments. The API handles over 100 million requests per day, generating annualized revenue of approximately 50 million dollars — remarkable traction for a company less than two years old.
The company's voice AI product, which enables natural conversation in Indian languages, has been deployed by several major call centers to handle customer service interactions. Early results show a 60% reduction in call handling time and a 40% improvement in customer satisfaction scores compared to traditional IVR systems. The product is particularly effective for financial services, where customers often prefer to discuss sensitive matters in their native language.
Government Partnership
The Indian government has been an early and enthusiastic supporter of Sarvam AI's mission. The Ministry of Electronics and Information Technology has signed a memorandum of understanding to use Sarvam's technology for several Digital India initiatives, including a multilingual AI assistant for government services that will be accessible via the Umang app used by over 50 million citizens.
Prime Minister Modi personally met with the Sarvam AI founding team at a technology summit, calling their work a critical step toward ensuring that the benefits of artificial intelligence reach every Indian, regardless of the language they speak. The government's AI Mission has allocated 500 crore rupees specifically for Indian language AI development, with Sarvam AI as a key beneficiary.
The Road Ahead
The 400 million dollar funding will be used to expand the engineering team from 150 to 600 employees, build proprietary GPU infrastructure to reduce dependence on cloud providers, expand language support to all 22 scheduled languages plus major regional dialects, and establish offices in Mumbai, Delhi, and Chennai. International expansion to Southeast Asian markets with similar linguistic diversity is planned for 2026.
Sarvam AI's success has inspired a wave of Indian AI startups focused on vernacular language applications. Investors are increasingly recognizing that the next billion AI users will not be English speakers, and that the companies building AI for non-English languages will capture enormous value. India, with its linguistic diversity and large population, is the most important battleground for this next phase of AI adoption.
